Graph engines
For many years, “I need a graph” meant one thing operationally: set up Neo4j, design something like a schema, build an ETL pipeline, and manage a server process. For many teams, that’s where graph projects went to die, not because graphs weren’t useful, but because the overhead arrived before the value. Too many vendors (including Neo4j) also focus(ed) on the data engineering part instead of creating business value.
Meanwhile, something has shifted quietly. The category of graph engine has emerged, and it’s worth understanding how it differs from a graph database. A graph database owns your data. It has native storage, indexing, a query language, and its own operational surface. You move your data in, and then you query it. That’s the right choice when graph is your primary data model and you need transactional guarantees at scale.
A graph engine is a query layer. Your data stays where it is (in DuckDB, Postgres, Iceberg, a data lake) and the engine gives it graph semantics on top. No migration. No ETL pipeline. No new infrastructure to babysit. MemGraph Zero goes even a step further and they position a graph database as an ETL solution with data translation on the fly.
This distinction matters more than it sounds. Most real-world graph use cases aren’t “we need a graph-native system.” They’re “we have data that has relationships, and we want to traverse those relationships.” Those are different requirements with very different solutions.
At the other end of the spectrum, tools like PuppyGraph take this further, they sit as a virtual graph layer over your existing warehouses and lakes, with zero data movement, querying Postgres, BigQuery, and Iceberg tables as a unified graph.
The question to ask before your next graph project isn’t “which graph database should I use?” It’s: do I need a database at all, or just a graph lens on data I already have? The answer shapes everything downstream: architecture, cost, adoption, and whether the project survives contact with production. The graph landscape has become a spectrum, full of diversity and real opportunity. But I also see customers increasingly paralyzed by the choices. Too many vendors, too much motion, no independent voice. Everyone is selling something. Very few people take the time to actually listen.
Very few people have an independent and object advice. Plus, not everyone has the budget to hire McKinsey or PWC. I try to guide companies in an honest, old fashioned consulting style. Talking and taking time. Effectively listening with patience and, well, experience.